LarrySequino Kubernetes Rollout Check
"Deployment succeeded" means the API accepted the object, not that the service works.
Procedure
- Check the rollout status and the replica counts together. Desired, current, updated, and available are four different numbers and only the last one matters.
- List pods and look at restart counts, not just phase. A pod in Running with nine restarts is a crash loop that happens to be between crashes.
- Read the events since the deploy started. Image pull failures, failed scheduling, and evicted pods all show up here and nowhere else.
- Check readiness probe failures separately from liveness. A failing readiness probe holds traffic back quietly; a failing liveness probe restarts the pod loudly.
- Compare the running image digest against the one that was meant to ship. A tag that did not move is the quietest failed deploy there is.
- Watch error rate and latency for one full traffic cycle before calling it done.
Verdict
Healthy, degraded, or roll back. Say which, and if it is roll back, give the command.